Search Results (22,521)

Hyperspectral imaging (HSI) provides rich spectral information and serves as a non-destructive technique for forensic stain analysis. Conventional approaches often exhibit degraded performance due to the high dimensionality and spectral redundancy inherent in hyperspectral data. To address this challenge, a hyperspectral dataset comprising nine beverage stains—papaya, coffee, pomegranate, orange, tea, wine, whisky, rum, and brandy—is developed. Building on this dataset, an ensemble framework that combines an optimized autoencoder (AE), channel-attention (CA)-enhanced one-dimensional convolutional neural networks (1D CNNs), and a Limited Memory Broyden–Fletcher–Goldfarb–Shanno (L-BFGS-B)-based weighted fusion strategy is proposed. The autoencoder learns compact latent representations from the 204-band hyperspectral vectors, reducing redundancy while preserving discriminative spectral features. CA emphasizes informative spectral bands and improves stain separability. Multiple 1D CNN models are trained using different latent dimensionalities, and their class probability outputs are fused through an optimized L-BFGS-B weighting scheme, where higher-performing models contribute more strongly to the final decision. Experimental results demonstrate classification accuracies of 96.54%, 97.19%, and 97.86% for the AE32 CA, AE64 CA, and AE128 CA models, respectively, with the optimized ensemble achieving an accuracy of 98.28%. Additionally, the time-dependent evolution of beverage stain reflectance is systematically analyzed using overlapped, normalized reflectance signatures acquired at time intervals of 0 min, 1 h, 2 h, 3 h, 4 h, and 5 h. The results confirm that AE-based latent compression, CA, and L-BFGS-B optimized ensemble fusion enhance hyperspectral beverage stain classification, providing an effective and extensible framework for forensic trace evidence analysis. Full article

Host-based intrusion detection systems (HIDSs) can address the limitations of network-based detection by analyzing system calls and other low-level events. Many existing benchmark datasets remain inadequate for evaluating modern attacks because they were built in outdated environments and cover only a limited set of attack behaviors. To address this gap, this study builds a TextRank-based preprocessing pipeline on the LID-DS 2021 dataset and compares five end-to-end pipelines: Random Forest (RF), Long Short-Term Memory (LSTM), Convolutional Neural Network(CNN) + LSTM, LSTM, Bidirectional LSTM (BiLSTM), and CNN + Bidirectional Gated Recurrent Unit (BiGRU). Of the 15 scenarios in the dataset, six multi-stage attacks were excluded, and three representative scenarios were selected based on attack-category coverage and suitability for single-chunk host-level detection. Within these three selected scenarios and same-scenario file-level splits, the deep learning pipelines achieved F1-scores of 0.90–0.94, whereas RF ranged from 0.55 to 0.63. Among the evaluated pipelines, CNN + BiGRU produced the strongest overall results. These findings indicate that, under this constrained evaluation setting, sequential deep learning pipelines can be effective for scenario-specific system-call-based HIDS; however, broader generalization to unseen attacks or to the full LID-DS 2021 scenario set remains unverified. Full article

The phenomenon of CO₂ corrosion of downhole tubing is widespread in oil and gas extraction. Currently, there is a lack of applicable prediction methods for the corrosion rates of different metal tubing in the liquid phase CO₂ environment. To address this issue, this paper systematically investigates the anti-corrosion mechanisms and influencing factors of different metal casings and proposes a deep learning model combining convolutional neural networks and long short-term memory networks. Based on laboratory corrosion experimental data, the model extracts spatial features of parameters affecting the corrosion rate through CNN and captures their temporal dependencies through LSTM. This paper builds a pipe corrosion rate prediction model based on the TensorFlow framework and compares the prediction results with those of the traditional D-W empirical model and the SRV machine learning model. The results showed that the CNN-LSTM model maintained high prediction accuracy regardless of high or low chromium content, with R² reaching 0.83 and 0.94 respectively, solving the problem that existing models have difficulty effectively simulating complex corrosion behavior under flowing corrosive media conditions. The model was verified using the remaining wall thickness of the actual application casing in the field, and the accuracy was over 80%. The established prediction method can be extended to predict the corrosion rate of pipes under similar corrosion conditions. Full article

Diabetic retinopathy (DR) is a leading cause of vision impairment and permanent blindness worldwide, requiring accurate and automated systems for multi-grade severity classification. However, standard Convolutional Neural Networks (CNNs) often struggle to capture fine, high-frequency microvascular patterns critical for diagnosis. This study proposes a Robust Intelligent CNN Model (RICNN) that integrates Gabor-based feature extraction with deep learning to improve DR classification. Specifically, Gabor filters are applied during preprocessing to extract orientation- and frequency-sensitive texture features, which are transformed into feature maps and concatenated with CNN feature representations at the fully connected layer (feature-level fusion). The model also incorporates the Synthetic Minority Oversampling Technique (SMOTE) for data balancing and the Adam optimizer for efficient convergence. This integration enhances sensitivity to microvascular structures such as microaneurysms and hemorrhages. The proposed RICNN was evaluated on the Messidor dataset (1200 images) across four severity levels: Mild, Moderate, Severe, and Proliferative DR. The model achieved an accuracy of 89%, a precision of 88.75%, a recall of 89%, and an F1-score of 89%, with AUCs of 97% for Severe DR and 99% for Proliferative DR. Comparative analysis confirms that the proposed texture-aware Gabor enhancement significantly outperforms LBP and Color Histogram approaches, indicating its potential for reliable clinical decision support. Full article

The detection and identification of unmanned aerial vehicles (UAVs) using radio frequency (RF) signals becomes particularly challenging in congested spectral environments, where conventional approaches relying solely on spectral characteristics often prove limited. This work introduces a novel technique for both UAV detection and classification based on temporal representations derived directly from the envelope of received RF signals. The proposed system follows a two-stage architecture: first, binary detection of UAV presence in a given RF channel, and second, identification of the specific UAV model among several commercial platforms. For the first stage, two signal representation methodologies are employed—Gramian Angular Fields and Hilbert curves—both generated from short-time RF windows and subsequently used as inputs to convolutional neural networks. Experimental evaluation demonstrates that the detection stage achieves accuracy rates exceeding 94% for the non-UAV class and approaching 99% for the UAV class with both approaches. In the identification stage, the system attains an accuracy above 90% for most considered UAV models, reaching up to 100% for certain platforms. These results confirm the effectiveness of the envelope-based approach for analyzing UAV-related RF signals. Full article

The automatic classification of ISUP (International Society of Urological Pathology) grade groups in prostate histopathological images remains challenging due to the high similarity between adjacent classes, class imbalance, and label noise. In this work, we propose a deep learning pipeline based on EfficientNet convolutional neural networks combined with a hybrid loss function that integrates ordinal regression and Focal Loss to better capture the ordered nature of ISUP grades. A noise-filtering strategy based on the entropy of predictions from multiple EfficientNet models was first applied to identify and remove high-uncertainty samples from the training set. The problem was then reformulated as an ordinal regression task to explicitly model the hierarchical relationship among grades. Experiments conducted on the PANDA dataset demonstrate that removing noisy samples improved performance from $\kappa =0.826$ to $\kappa =0.833$. Incorporating ordinal loss further increased performance to $\kappa =0.851$. The best configuration, combining ordinal regression and Focal Loss, achieved $\kappa =0.857$ and an accuracy of $0.669$, while reducing severe misclassifications and concentrating errors among adjacent classes. These results indicate that explicitly modeling ordinal structure and mitigating label noise are effective strategies for improving prostate cancer grading systems. Full article

Spiking neural networks (SNNs) offer inherent advantages in processing temporal information. However, their network topologies are predominantly algorithm-generated, lacking constraints from biological brain connectivity, which limits their bio-plausibility. In our previous work, we constructed a spiking neural network (SNN) by incorporating the topological structure of functional brain networks derived from fMRI data of healthy subjects and proposed an fMRISNN model. This model was further employed as the reservoir layer of a liquid state machine (LSM) to build a speech recognition framework. In this framework, the Lyon ear model and the BSA were used to encode speech signals into spike sequences; however, this approach suffers from high computational cost and limited adaptability to temporal variations. To address these limitations, we propose an enhanced Mel-frequency cepstral coefficient (MFCC)-driven sparse spike encoding method. For the speech recognition task, we systematically compare the two preprocessing pipelines in terms of spike number, spike sparsity, encoding time, and downstream speech recognition performance. Experimental results show that the proposed method generates substantially fewer spikes, achieves markedly higher sparsity, and requires significantly less encoding time, while maintaining nearly the same recognition accuracy under the same LSM-based framework. These findings indicate that improved speech input representation can enhance the computational efficiency of SNN-based speech recognition without compromising recognition capability. In addition, the fMRISNN model significantly outperforms several baseline models with algorithmically generated topologies. Compared with mainstream models reported in the literature, although the deep convolutional neural network (CNN) still achieves higher absolute recognition accuracy, the fMRISNN exhibits clear advantages in terms of model parameter size and theoretical energy efficiency. Full article

Quantum error-correcting codes are essential for achieving fault-tolerant quantum computing. Heavy hexagonal code is a type of topological code that leverages the arrangement of qubits to find and correct errors. The heavy hexagonal code is suitable for superconducting architectures, specifically graph layouts with a limited number of connections. The topological error correction methods work well, but they need more qubits, cannot be used for different sizes of quantum systems, are less reliable, and do not work well with changing quantum distributions. Thus, the research proposes an Ardea-guided exploit optimized sparse-dual attention enabled meta-learning-based convolutional neural network with bi-directional long short-term memory model (AGuESD-MCBiTM). The method exhibits effective correction over dynamic environments with the utilization of meta-learning and the extraction of statistical information, which provides a detailed representation of the qubit patterns. The Ardea-guided exploit optimized (AGuEO) algorithm tunes the weights of MCBiTM and acquires optimal solutions with higher convergence. Moreover, the sparse-dual attention module and meta-learning-based MCBiTM model, which together provide scalable, real-time identification of non-linear qubit noise fluctuations with lower computational cost. Comparatively, the proposed AGuESD-MCBiTM exhibits superior error correction with a higher correlation of 0.97, accuracy of 98.93%, and R-squared value of 0.93, as well as a lower Root mean square error of 1.87, Mean absolute error of 1.20, Bit error rate of 1.85, Logical error rate of 3.82, and mean square error of 3.49 in circuit 2, respectively. Full article

With the development of geological data analysis and big data technology, intelligent mineral prospectivity mapping (MPM) has become a key direction in the integration of geoscience and artificial intelligence, showing promising applications in the identification and evaluation of strategic mineral resources such as gold. To address the limitations of conventional methods—including insufficient training samples, complex model structures, and weak capability in recognizing anomalous zones—this study proposes an improved convolutional neural network (CNN) approach for mineral prediction. A lightweight, modular CNN structure with repeatable stacking is designed to reduce computational cost while enhancing model robustness and generalization. In addition, a dynamic learning rate scheduling strategy is adopted to optimize the training process, significantly improving convergence speed and training stability. Furthermore, high-probability prediction samples and low-probability background samples are combined to form a new training dataset for regional prospectivity evaluation, yielding a high area under the curve (AUC) score. The method is applied and validated in the Xiong’ershan region, and the predicted high-potential zones account for 30% of the study area and contain 81.4% of the known gold deposits. These results demonstrate the method’s effectiveness in mineral information extraction and blind-area targeting, offering a new approach for mineral prospectivity mapping. Full article

Cloud detection is essential for quantitative land-surface remote sensing and cloud-climate research. However, existing methods often prioritize spatial features over spectral features, which limits thin-cloud detection. To address this issue, this paper proposes a Thin-Cloud-Sensitive Network (TCSNet) for hyperspectral imagery. TCSNet employs an encoder–decoder architecture with a dual-branch design: a convolutional neural network (CNN) extracts multi-scale local features, while a PVTv2-B2 Transformer captures long-range spectral dependencies. To effectively integrate the complementary representations from both branches, a Cross-Modal Fusion (CMF) module with a lightweight single-channel gate is introduced at each stage, followed by a channel attention mechanism (SE) for feature recalibration. Subsequently, a Multi-Scale Fusion (MSF) module is used to integrate multi-level features through a top-down pathway, enabling deep semantic information to guide shallow feature expression. Furthermore, to enhance the decoder’s feature representation capability, a Combined Attention Mechanism (CAM) is incorporated at each decoder stage. This design enables the network to simultaneously focus on important channels, salient regions, and cloud boundaries, effectively alleviating spectral confusion between thin clouds and the underlying surface. Experimental results on Gaofen-5 01 hyperspectral data demonstrate that TCSNet achieves the highest recall (92.98%), ${\mathrm{R}\mathrm{e}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{l}}_{thin}$ (85.59%), and ${\mathrm{R}\mathrm{e}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{l}}_{thick}$ (99.75%), thereby validating its superiority for thin-cloud detection. Full article

Aiming at the issues of low recognition accuracy and high model computational complexity for power quality disturbances (PQDs) in strong-noise environments, this paper proposes a novel lightweight PQD-recognition method that integrates a hybrid architecture of variational mode decomposition (VMD), convolutional neural network (CNN), and transformer. Firstly, a hybrid optimization algorithm named the monkey–genetic hybrid optimization algorithm (MGHOA) is proposed to optimize VMD parameters for denoising disturbance signals, thereby enhancing recognition accuracy in noisy environments. Secondly, to fully extract disturbance signal features and reduce the computational complexity of the model, a lightweight CNN–transformer model is designed. Depthwise separable convolution (DSC) is employed to extract local features and the multi-head attention mechanism of transformer is utilized to mine the long-distance dependence and global features, thereby enhancing the feature representation. Thirdly, a multitask joint-learning method is proposed to collaboratively optimize classification accuracy and temporal localization tasks, enhancing the discrimination of similar disturbances. Additionally, a dual-pooling global feature fusion strategy is designed to further enhance the model’s ability to discriminate complex disturbances. Comparative experiments on 16 typical PQD types demonstrate that the proposed method achieves excellent performance in recognition accuracy, model robustness, and computational efficiency. The integration of the MGHOA–VMD module improves recognition accuracy by 1.08%, while the multitask joint-learning method contributes an additional 0.55% improvement. When achieving recognition accuracy comparable to complex models, the training time of the proposed method is 36.51% of that required by DeepCNN and merely 5.90% of that required by bidirectional long short-term memory (BiLSTM), with a 31.22% reduction in parameter scale. This work provides a novel solution for intelligent power quality disturbance recognition. Full article

Voice disorders severely limit verbal communication, creating a need for intuitive assistive technologies. To meet this need, we present epidermal strain sensors that capture strain signals during silent speech and hand gesture. A thin electrospun nanofiber layer integrated onto commercial polyurethane films guides uniform, controlled microcrack formation in screen-printed carbon conductive paths, achieving a gauge factor up to 243 over 0–40% strain. Signals from the seven-channel strain sensor array are recognized by a hybrid neural network that combines convolutional and Transformer architectures, reaching over 98% accuracy. The recognized outputs are rendered in virtual reality (VR), enabling intuitive, real-time communication. Moreover, the approach simplifies fabrication by enabling crack-based strain sensing with only a thin electrospun surface layer on commercial polyurethane films, eliminating the need for thick freestanding electrospun substrates. This cost-effective approach addresses limitations of conventional electrospun substrates by minimizing the thickness of the electrospun layer, thereby shortening the electrospinning time. Overall, the work demonstrates a method for translating natural non-verbal expressions into speech and text in VR, with promising applications in healthcare and assistive communication. Full article

The increasing complexity of next-generation wireless systems challenges the scalability and generalization capabilities of traditional model-driven physical layer (PHY) design, which relies on analytically derived channel models and optimization frameworks. This paper presents a comprehensive survey and critical review of deep learning (DL) architectures enabling the transition toward AI-native PHY design. A unified optimization perspective is developed in which all PHY tasks—including channel estimation, channel state information (CSI) feedback, massive MIMO processing, signal detection, channel coding, beamforming, resource allocation, and semantic-aware transmission—are formulated under a common empirical risk minimization (ERM) framework. Neural architectures such as autoencoders, convolutional and recurrent networks, transformers, and reinforcement learning models are examined through their underlying optimization formulations, loss functions, training methodologies, and representation learning mechanisms. The review compares model-driven and AI-native approaches in terms of performance metrics, computational complexity, robustness, generalization capability, and practical deployment constraints, including hardware limitations, energy efficiency, and real-time feasibility. The analysis highlights the conditions under which AI-native architectures provide adaptability and performance improvements while identifying trade-offs in complexity, latency, and interpretability. The study concludes by outlining prioritized research directions toward fully adaptive and self-optimizing wireless communication systems. Full article

Most neural architectures model time as a one-dimensional real-valued variable, constraining temporal reasoning to sequential propagation along a single axis. We introduce Complex-Time Neural Networks (CTNN), a new class of architectures in which temporal coordinates are elements of the complex plane T = t + iτ ∈ ℂ, where Re(T) preserves chronological ordering and Im(T) encodes an orthogonal experiential dimension. Within this geometry, Im(T) < 0 defines a memory domain enabling retrospective retrieval, Im(T) = 0 corresponds to present-moment computation, and Im(T) > 0 defines an imagination domain for prospective projection. We prove the Expressive Separation Theorem (Theorem 1), establishing that, within the temporally coupled function class G_TCP and under explicit Assumptions A1–A4 (in particular the bounded projection Assumption A3), CTNN accesses temporally coupled functions at O(1) cost with respect to temporal distance Δ₁, Δ₂, while real-time architectures incur Ω(Δ₁ + Δ₂) sequential steps. For layered compositions, this yields an exponential composition gap within G_TCP under A1–A4. These advantages hold under the stated assumptions and may not directly generalize to broader function classes or large-scale settings where A3 cannot be maintained. Therefore, Theorem 1 provides a formal separation result for G_TCP, while CTNN more broadly defines a geometric framework for temporal computation. As the first concrete instantiation of this framework, we develop Complex-Time Convolutional Neural Networks (CTCNN). CTCNN achieves state-of-the-art performance on Something-Something V2 (70.2 ± 0.4%, +1.1% over VideoMAE v2, p < 0.01), strong performance on Kinetics-400 (78.4 ± 0.3%), and substantial gains on Long Range Arena Path-X (87.3% vs. 79.6%, +7.7%), using 3.4× fewer parameters than VideoMAE v2. Learnable angular parameters α and β provide computationally interpretable parameters related to memory-access span and prospection breadth, with values varying systematically across task families. Full article

A multimodal strategy for fall detection within the broader domain of human activity recognition is developed in this study. A fine-tuned Inflated 3D Convolutional Network model, trained in optical flow data derived from video inputs, achieves 92.70% accuracy in classifying fall-related events. Simultaneously, a Convolutional Neural Network–Bidirectional Long Short-Term Memory model incorporating attention mechanisms processes time-series sensor data, contributing to an ensemble performance of 97.87%. The integration of visual and sensor modalities illustrates a promising direction for developing reliable, real-time fall detection systems applicable in healthcare and assisted living environments. Full article